Chase Granberry

Interesting. Thank you for the detail here! Maybe something in the wire protocol causes more round trips for citext tables. We don't do anything specific for any types.

With lots of tenants we can only have so many (very few) metrics before the endpoint Prometheus scrapes gets HUGE. e.g. the Realtime endpoint is like 10MB at the moment...

@janpio no, this is still re: implementing Otel tracing along the query pipeline. We have already implemented a lot of metrics which are [exposed](https://github.com/supabase/supavisor/blob/4265b5143a4577f064b5fb1d7b23c6cd6bc36c18/lib/supavisor_web/router.ex#L51) at `/metrics` (for the whole system)...

Can't replicate:  The log for bad password: ``` ClientHandler: Exchange error: "Wrong password" when method :auth_query ```

And when I try a user which does not exist:  And the log message: ``` ClientHandler: Authentication auth_secrets error: "There is no user 'joe' in the database. Please create...

Closing for now...

Updated and re-opened

Needs support on the Postgres side as well. Unsure if there is much benefit until then. Maybe for fail2ban? [Latest Postgres patch thread](https://github.com/supabase/supavisor/issues/204)

> I pushed an edit to rewrite some bits for more clarity. Very much appreciated @charislam

What is "large" exactly here?